There are loads of mind-blowing advancements being made in robotics and artificial intelligence, and they’re bringing us a whole lot of cool stuff. Self-driving cars, delivery drones, at least one robot-run hotel…the list goes on.
What has people most excited, though, are the advances in artificial intelligence and processing power, both of which are necessary to build robots that can think, move, and respond almost like real human beings. Some of today’s robots can recognize emotions, hold basic conversations, and even crack a few jokes (though by all appearances those jokes are pre-written and pre-programmed – when AI starts coming up with its own puns, I’ll be out of a job).
I have to confess, though, I’m left with a giant question when it comes to these androids: Why?
A look at the current generation of humanoid robots and at the complexity of human interactions with human-like technology makes me think today’s robots are built more on short-term novelty than on shocking innovation…with a healthy dose of creepiness that will keep them from appealing to the masses.
Humanoid Robots Suck as Servants
Japan’s Henn-na Hotel has been a bit of a media darling since its opening. Not because it offers great luxury or award-winning service but because it is staffed almost entirely by robots. An automated mannequin approximating the look of a Japanese female greets you at the front desk if you speak Japanese. Gaijin, like myself, have to make do with the English-speaking animatronic T-rex.
There’s an automated trolley that takes your bags to your room for you, a robotic arm that stores and retrieves secure boxes to replace outdated hotel safes, and door locks are operated by digital facial recognition rather than keycards.
It’s the perfect use of robotics, and there’s hardly a humanoid feature to it. That’s not a coincidence.
The only human-ish robots you’ll meet at the hotel are the two stuck behind the front desk, and they are purposefully gimmicky; they are a fun way to introduce you to the concept, not an attempt to force you into a true interaction with a machine.
The benefits and drawbacks (so far) of an automated hotel are fairly obvious: cost savings, consistency, no questions about tips for foreign travelers, no real concierge service, no asking the staff what their favorite local eateries are, etc. The benefits of achieving this automation through robots attempting to appear human, however, are absolutely nil.
Then there’s Jibo.
They haven’t shipped any units, though they report “selling” 4,800 Jibos through their 2,000%+ funded IndieGogo campaign. By all appearances, they’re poised to take the in-home robot market by storm.
So what does this fantastically well-funded ($40MM+ according to what we know) robot do? It sits on your counter or tabletop and uses a nice, human voice to remind you about appointments, help you make phone and video calls, and provide some very basic interaction that sort of, kind of, almost feels human.
Jibo’s uniqueness is in its voice, its personality, its ability to spin around and swivel cutely, and in the “face” it displays on its screen. Basically, it’s a stationary smartphone / virtual assistant (Siri 2.0?) designed to look more adorable and encourage engagement.
More human, does less. Like I said, the novelty is clear, but once that wears off, how many people are going to remain thrilled with their $749 robot?
Maybe they’ll want to step up to Pepper, the Japan-designed Taiwan-made robot that has now sold out 1,000-unit runs over five successive generations. At 198,000 yen (around $1,700), this robot costs more than this writer’s first car, and is just about as functional (just kidding; I still miss my ’93 Volkswagen Golf).
Pepper’s only purpose is to be a companion. Even though it has hands and arms and can move around on flat surfaces, it can’t actually do anything. It can’t make a sandwich, make a phone call, or even fold laundry (though newcomer Laundroid apparently can as “The World’s First Laundry Folding Bot”). Pepper’s AI algorithms enable it to sense and respond to emotion through some vocal cues, facial expressions, and body language, which is impressive (or will be when it gets better at it), but it’s not nearly humanoid enough to fool anyone into think they’re having an actual interpersonal interaction.
Even more human, does even less.
When it comes to actual functionality – by which I mean the ability to do things in the physical world, thus saving human energy and time – robots don’t need to look or act like humans at all. This has long been known in industrial settings, with bomb-deactivating robots, with the Mars Rover and other vehicles. Human bodies are decent at many things, but designing machines to be good at specific purposes makes for more efficient operation and manufacturing.
So Pepper’s makers might actually have the right idea insofar as designing their robot purely for companionship rather than task-completion, but realistically, these kinds of androids are going to be lacking for a while due to an inherent creepiness many feel when it comes to humanoid robots.
Compounding Lost Functionality With the Uncanny Valley
The hypothesis of the Uncanny Valley, which has been borne out in robotics and AI research, states that people (at least some) will respond ever more positively and with more empathy to robots and other machines/interfaces as they appear more and more human. That is, until a certain point of humanness is reached, at which point feelings instantly flip to discomfort and revulsion.
When a robot appears to be almost human, but not quite, we’re more likely to fear and hate them than we are to want them as companions or servants.
The fact that the Henn-na Hotel doesn’t have humanoid robots is actually a good thing – emotionally and psychologically. The fact that Pepper, and especially Jibo don’t look or act fully humanoid also scores them points, at least in terms of not severely weirding anyone out.
But there’s the Catch-22.
If these “social robots” or companion robots are meant to provide human interaction that is otherwise missing, filling the primal need we have for interhuman connectedness, then they need to be convincing facsimiles of real people. But unless they are perfect facsimiles of people – so human-like that real people can’t tell they’re robots – they’ll turn off the majority of their would-be companions.
Right now, robot makers have to decide between making functional robots that actually improve efficiency and productivity in our lives, or human companion robots that earn full marks for novelty but will end up gathering dust in a closet six months after purchase.
If and when technology reaches the point where we can build robots indistinguishable from humans, they might make for better companions. Of course at that level of emotion and creative thinking opens up the ethical doors presaged by decades of science fiction; a stage liable to make people at least as uncomfortable as the Uncanny Valley.
We don’t need more people. There are already more than 7 billion of those, and they’re pretty good at reproducing (with the not-so-coincidental exception of Japan). Maybe we should focus on building robots that do things better than humans can do them, and let people do the peopling.
